Generative Neural Networks for Anomaly Detection in Crowded Scenes (2019-05-01T00:00:00.000000Z)

Abstract

Security surveillance is critical to social harmony and people’s peaceful life. It has a great impact on strengthening social stability and life safeguarding. Detecting anomaly timely, effectively and efficiently in video surveillance remains challenging. This paper proposes a new approach, called <inline-formula> <tex-math notation="LaTeX">$S^{2}$ </tex-math></inline-formula>-VAE, for anomaly detection from video data. The <inline-formula> <tex-math notation="LaTeX">$S^{2}$ </tex-math></inline-formula>-VAE consists of two proposed neural networks: a Stacked Fully Connected Variational AutoEncoder (<inline-formula> <tex-math notation="LaTeX">$S_{F}$ </tex-math></inline-formula>-VAE) and a Skip Convolutional VAE (<inline-formula> <tex-math notation="LaTeX">$S_{C}$ </tex-math></inline-formula>-VAE). The <inline-formula> <tex-math notation="LaTeX">$S_{F}$ </tex-math></inline-formula>-VAE is a shallow generative network to obtain a model like Gaussian mixture to fit the distribution of the actual data. The <inline-formula> <tex-math notation="LaTeX">$S_{C}$ </tex-math></inline-formula>-VAE, as a key component of <inline-formula> <tex-math notation="LaTeX">$S^{2}$ </tex-math></inline-formula>-VAE, is a deep generative network to take advantages of CNN, VAE and skip connections. Both <inline-formula> <tex-math notation="LaTeX">$S_{F}$ </tex-math></inline-formula>-VAE and <inline-formula> <tex-math notation="LaTeX">$S_{C}$ </tex-math></inline-formula>-VAE are efficient and effective generative networks and they can achieve better performance for detecting both local abnormal events and global abnormal events. The proposed <inline-formula> <tex-math notation="LaTeX">$S^{2}$ </tex-math></inline-formula>-VAE is evaluated using four public datasets. The experimental results show that the <inline-formula> <tex-math notation="LaTeX">$S^{2}$ </tex-math></inline-formula>-VAE outperforms the state-of-the-art algorithms. The code is available publicly at <uri>https://github.com/tianwangbuaa/</uri>.